2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 #include "xfs_types.h"
24 #include "xfs_trans.h"
29 #include "xfs_dmapi.h"
30 #include "xfs_mount.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_alloc_btree.h"
33 #include "xfs_ialloc_btree.h"
34 #include "xfs_dir_sf.h"
35 #include "xfs_dir2_sf.h"
36 #include "xfs_attr_sf.h"
37 #include "xfs_dinode.h"
38 #include "xfs_inode.h"
39 #include "xfs_inode_item.h"
41 #include "xfs_btree.h"
42 #include "xfs_ialloc.h"
43 #include "xfs_itable.h"
44 #include "xfs_dfrag.h"
45 #include "xfs_error.h"
52 * Syssgi interface for swapext
56 xfs_swapext_t __user *sxu)
59 xfs_inode_t *ip=NULL, *tip=NULL;
61 xfs_vnode_t *vp = NULL, *tvp = NULL;
62 struct vnode *bvp, *btvp;
65 sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
67 error = XFS_ERROR(ENOMEM);
76 if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
77 error = XFS_ERROR(EFAULT);
81 /* Pull information for the target fd */
82 if (fgetvp(td, (int)sxp->sx_fdtarget, &bvp) != 0) {
83 error = XFS_ERROR(EINVAL);
90 error = XFS_ERROR(EBADF);
94 if (fgetvp(td, (int)sxp->sx_fdtmp, &btvp) != 0) {
95 error = XFS_ERROR(EINVAL);
99 tvp = VPTOXFSVP(btvp);
102 error = XFS_ERROR(EBADF);
106 if (ip->i_mount != tip->i_mount) {
107 error = XFS_ERROR(EINVAL);
111 if (ip->i_ino == tip->i_ino) {
112 error = XFS_ERROR(EINVAL);
118 if (XFS_FORCED_SHUTDOWN(mp)) {
119 error = XFS_ERROR(EIO);
123 error = XFS_SWAP_EXTENTS(mp, &ip->i_iocore, &tip->i_iocore, sxp);
134 kmem_free(sxp, sizeof(xfs_swapext_t));
148 xfs_bstat_t *sbp = &sxp->sx_stat;
149 xfs_vnode_t *vp, *tvp;
150 xfs_ifork_t *tempifp, *ifp, *tifp;
151 int ilf_fields, tilf_fields;
152 static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
161 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
163 error = XFS_ERROR(ENOMEM);
171 /* Lock in i_ino order */
172 if (ip->i_ino < tip->i_ino) {
180 xfs_lock_inodes(ips, 2, 0, lock_flags);
183 /* Check permissions */
184 error = xfs_iaccess(ip, S_IWUSR, NULL);
188 error = xfs_iaccess(tip, S_IWUSR, NULL);
192 /* Verify that both files have the same format */
193 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
194 error = XFS_ERROR(EINVAL);
198 /* Verify both files are either real-time or non-realtime */
199 if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
200 (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
201 error = XFS_ERROR(EINVAL);
205 /* Should never get a local format */
206 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
207 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
208 error = XFS_ERROR(EINVAL);
212 if (VN_CACHED(tvp) != 0) {
213 xfs_inval_cached_trace(&tip->i_iocore, 0, -1, 0, -1);
214 XVOP_FLUSHINVAL_PAGES(tvp, 0, -1, FI_REMAPF_LOCKED);
217 /* Verify O_DIRECT for ftmp */
218 if (VN_CACHED(tvp) != 0) {
219 error = XFS_ERROR(EINVAL);
223 /* Verify all data are being swapped */
224 if (sxp->sx_offset != 0 ||
225 sxp->sx_length != ip->i_d.di_size ||
226 sxp->sx_length != tip->i_d.di_size) {
227 error = XFS_ERROR(EFAULT);
232 * If the target has extended attributes, the tmp file
233 * must also in order to ensure the correct data fork
236 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
237 error = XFS_ERROR(EINVAL);
242 * Compare the current change & modify times with that
243 * passed in. If they differ, we abort this swap.
244 * This is the mechanism used to ensure the calling
245 * process that the file was not changed out from
248 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
249 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
250 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
251 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
252 error = XFS_ERROR(EBUSY);
256 /* We need to fail if the file is memory mapped. Once we have tossed
257 * all existing pages, the page fault will have no option
258 * but to go to the filesystem for pages. By making the page fault call
259 * VOP_READ (or write in the case of autogrow) they block on the iolock
260 * until we have switched the extents.
263 error = XFS_ERROR(EBUSY);
267 xfs_iunlock(ip, XFS_ILOCK_EXCL);
268 xfs_iunlock(tip, XFS_ILOCK_EXCL);
271 * There is a race condition here since we gave up the
272 * ilock. However, the data fork will not change since
273 * we have the iolock (locked for truncation too) so we
274 * are safe. We don't really care if non-io related
278 XVOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
280 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
281 if ((error = xfs_trans_reserve(tp, 0,
282 XFS_ICHANGE_LOG_RES(mp), 0,
284 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
285 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
286 xfs_trans_cancel(tp, 0);
290 xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
293 * Count the number of extended attribute blocks
295 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
296 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
297 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
299 xfs_trans_cancel(tp, 0);
303 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
304 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
305 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
308 xfs_trans_cancel(tp, 0);
314 * Swap the data forks of the inodes
318 *tempifp = *ifp; /* struct copy */
319 *ifp = *tifp; /* struct copy */
320 *tifp = *tempifp; /* struct copy */
323 * Fix the on-disk inode values
325 tmp = (__uint64_t)ip->i_d.di_nblocks;
326 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
327 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
329 tmp = (__uint64_t) ip->i_d.di_nextents;
330 ip->i_d.di_nextents = tip->i_d.di_nextents;
331 tip->i_d.di_nextents = tmp;
333 tmp = (__uint64_t) ip->i_d.di_format;
334 ip->i_d.di_format = tip->i_d.di_format;
335 tip->i_d.di_format = tmp;
337 ilf_fields = XFS_ILOG_CORE;
339 switch(ip->i_d.di_format) {
340 case XFS_DINODE_FMT_EXTENTS:
341 /* If the extents fit in the inode, fix the
342 * pointer. Otherwise it's already NULL or
343 * pointing to the extent.
345 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
346 ifp->if_u1.if_extents =
347 ifp->if_u2.if_inline_ext;
349 ilf_fields |= XFS_ILOG_DEXT;
351 case XFS_DINODE_FMT_BTREE:
352 ilf_fields |= XFS_ILOG_DBROOT;
356 tilf_fields = XFS_ILOG_CORE;
358 switch(tip->i_d.di_format) {
359 case XFS_DINODE_FMT_EXTENTS:
360 /* If the extents fit in the inode, fix the
361 * pointer. Otherwise it's already NULL or
362 * pointing to the extent.
364 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
365 tifp->if_u1.if_extents =
366 tifp->if_u2.if_inline_ext;
368 tilf_fields |= XFS_ILOG_DEXT;
370 case XFS_DINODE_FMT_BTREE:
371 tilf_fields |= XFS_ILOG_DBROOT;
375 #ifdef XXXKAN /* Not necessary, vnodes are vrefed already by fgetvp */
377 * Increment vnode ref counts since xfs_trans_commit &
378 * xfs_trans_cancel will both unlock the inodes and
379 * decrement the associated ref counts.
385 xfs_trans_ijoin(tp, ip, lock_flags);
386 xfs_trans_ijoin(tp, tip, lock_flags);
388 xfs_trans_log_inode(tp, ip, ilf_fields);
389 xfs_trans_log_inode(tp, tip, tilf_fields);
392 * If this is a synchronous mount, make sure that the
393 * transaction goes to disk before returning to the user.
395 if (mp->m_flags & XFS_MOUNT_WSYNC) {
396 xfs_trans_set_sync(tp);
399 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
405 xfs_iunlock(ip, lock_flags);
406 xfs_iunlock(tip, lock_flags);
409 kmem_free(tempifp, sizeof(xfs_ifork_t));